Branch: refs/heads/thaines/AbsRegionConverter_convert
  Home:   https://github.com/dyninst/dyninst
  Commit: 808f3dd21e5351d489007f7952466cb2dd6c4469
      
https://github.com/dyninst/dyninst/commit/808f3dd21e5351d489007f7952466cb2dd6c4469
  Author: Tim Haines <thaines.as...@gmail.com>
  Date:   2025-01-15 (Wed, 15 Jan 2025)

  Changed paths:
    M dataflowAPI/src/AbslocInterface.C

  Log Message:
  -----------
  Preserve full register in AbsRegionConverter::convert(RegisterAST::Ptr)

ROSE distinguishes partial registers from full ones, so that needs to be
reflected here since this class is the main point of conversion between
Dyninst and ROSE.

Example from aarch64:

  0x94C: ldr W0, [X1, W0, lsl 0x2]

The created Assignment is

  @94c<[aarch64::x0]>[aarch64::x1]>[aarch64::x0]>H[]

Note that the assignment thinks the first register is x0, rather than
w0. Passing this through AssignmentConverter::convert, the generated
ROSE expressions are

  SgAsmIntegerType:SgAsmDirectRegisterExpression for 0x94C W0
  SgAsmIntegerType:SgAsmDirectRegisterExpression for 0x94C X1
  SgAsmIntegerType:SgAsmDirectRegisterExpression for 0x94C W0

Now, the first operand is correct. However, this operand is not
recognized in

    RegisterStateAST::writeRegister(
      const RegisterDescriptor&,
      const BaseSemantics::SValuePtr&,
      Dyninst::DataflowAPI::Result_t&,
      std::map<Dyninst::Absloc, Dyninst::Assignment::Ptr>&
    )

because the map was created using the data from the Assignment. This
causes the instruction expansion in SymbolicExpansion::expandAarch64
to fail.

The other option would be to check for base register equality in
Absloc::operator==, but that would change the semantics of equivalent
abstract locations.



To unsubscribe from these emails, change your notification settings at 
https://github.com/dyninst/dyninst/settings/notifications
_______________________________________________
Dyninst-api mailing list
Dyninst-api@cs.wisc.edu
https://lists.cs.wisc.edu/mailman/listinfo/dyninst-api

Reply via email to